halosimr

Simulate exposure to high-acuity, low-occurrence (HALO) events across a provider population, model how readiness decays between exposures, and estimate how training programs change it. An R implementation extending the methodology of the Code Blue Blindspots paper (PMID 41633464).

Installation

# install.packages("remotes")
remotes::install_github("EmergencyMind/halosimr")

Plotting uses ggplot2, which is optional:

install.packages("ggplot2")

Quick start

library(halosimr)

# 1. Generate a schedule and a stream of HALO events
sched <- generate_schedule(n_providers = 200, n_days = 365,
                           schedule_type = "3/7 Day", seed = 42)
ev    <- generate_events(n_days = 365, rate = 0.1, seed = 42)

# 2. Run one simulation
sim <- halo_sim(sched, ev, readiness_model = "exponential",
                readiness_half_life_days = 60)
sim <- hsrun(sim)
sim

plot_readiness(sim)
plot_gap_distribution(sim)

Repeat over many random draws to see the spread, and compare a training program against baseline:

mc <- hsrunmc(
  n = 50,
  schedule_fn = function(s) generate_schedule(200, 365, "3/7 Day", seed = s),
  events_fn   = function(s) generate_events(365, rate = 0.1, seed = s),
  readiness_model = "exponential", readiness_half_life_days = 60,
  training_program = "monthly"
)
mc
plot_mc_band(mc)

Readiness models

Four decay models map days since the last exposure or training reset to a readiness value in the range 0 to 1:

• binary: ready within a threshold, otherwise not
• exponential: halves every readiness_half_life_days
• ebbinghaus: forgetting curve 1 / (1 + b * days)
• step: full, then partial, then zero across two thresholds

Documentation

vignette("halosimr")

License

MIT (c) Daniel Dworkis. See LICENSE.